IOT & WIRELESS SENSOR NETWORKS 21EC741

IOT & WIRELESS SENSOR NETWORKS

(21EC741)
AS PER 2021 SCHEME
VISVESVARAYA TECHNOLOGICAL UNIVERSITY, BELAGAVI
NEP, Outcome Based Education (OBE) and Choice Based Credit System (CBCS)

Module-5
Middleware for WSNs:
Introduction
principles
Architecture
data related functions
Performance and traffic management:
background
WSN Design issues
performance modelling of WSNs.

Divya T M, Assistant Professor, Dept.of ECE

 IOT & WIRELESS SENSOR NETWORKS 21EC741

Middleware For Wireless Sensor Networks

Introduction :
WSNs are constrained in resources such as bandwidth, computation and
communication capabilities, and energy. WSN topology is variable due to node
mobility, depletion of energy, switching between sleep and active states, radio
range, and routing possibilities. WSN is a wireless/mobile and resource-
constrained network with diverse applications. Designing middleware for WSN is a
challenging task.
Middleware exists usually below the application level and on top of the operating
systems and network protocols. It marshals the application requirements, hides
details of lower levels, and facilitates application development and deployment and
their management. middleware helps negotiate between applications and low-level
network protocols in order to improve performance and save network resources.
the middleware needs to know the features of both applications and network
protocols. Middleware constructs an effective mapping between applications and
network protocol based on the current network status and the required application
QoS. This mapping may be implemented as middleware services that can be
invoked by applications. Middleware services provide application knowledge and
its QoS, as well as current network state, controls and manages network
resources.
WSN Middleware Principles/Challenges
The Challenges in the design of middleware for WSNs are :
(1) The topology control, to rearrange the sensor nodes into a connected network.
(2) The energy-aware data-centric computation.
(3) The application-specific integration, since integration of application information
into the network protocol improves performance and conserves energy.
(4) The efficient utilization of computational and communications resources.
(5) The support for real-time applications.
Functions of Middleware for WSN
1. middleware needs to provide a standardized system service because to
deploy current and future applications easily.
2. An environment that coordinates and supports multiple applications; this is
required to implement the diverse applications and to create new ones.
3. Mechanisms to achieve adaptive and efficient utilization of system resources;
these mechanisms provide algorithms that dynamically manage limited and
variable network resources of WSNs
4. Efficient trade-offs between the multiple QoS dimensions; this can be used
to adjust and optimize the required network resources.

Divya T M, Assistant Professor, Dept.of ECE

 IOT & WIRELESS SENSOR NETWORKS 21EC741

Design Principles of WSN Middleware/Need of WSN Middleware

1. need for localized algorithms as distributed algorithms that achieve a global
goal by communicating with nodes in some neighborhood
2. need for adaptive fidelity algorithms to trade-off between the quality of the
results and resource utilization
3. need for data-centric mechanisms for data processing and querying within
the network and for decoupling data from the physical sensor
4. need for Application knowledge, integrated into the services provided by the
middleware, to improve resource and energy efficiencies
5. need for lightweight middleware for both computation and communication;
(6) and need to perform application QoS trade-offs since given the limited
resources in WSNs.
6. need to perform application QoS trade-offs since given the limited resources
in WSNs.

MIDDLEWARE ARCHITECTURE
The middleware gathers information from the application and network protocols
and determines how to support the applications and adjusts network protocol
parameters.
The middleware interfaces with the operating system directly while bypassing the
network protocol.
The middleware consists of functional elements such as
 Resource management : The resource management functional element
monitors the network status and receives application requirements. It then
produces the command to adjust the network resource.
 Event detection & management : The event detection and management
functional element is used to detect and manage events such as sensing.
 Application programming interface (API) : The API can be invoked by
applications to achieve better performance and network utilization.

Divya T M, Assistant Professor, Dept.of ECE

 IOT & WIRELESS SENSOR NETWORKS 21EC741

Fig: General Middleware architecture for WSN

For example, QueryAgent defines a general programming interface between
application and lower layers. QueryAgent consists of a subset of APIs for
unicast communication.
The queryAgent works as follows:
Unicast() : The application sends this message to the general interface and
the interface returns a suitable routing protocol for the query.
API start_unicast() : this message is called and this is sent to where the query
is processed.
API listen_unicast(): this message is called in order to wait for the data from
the sensor network.
finish_unicast() : the application calls this message and receives the query
results.

QueryAgent consists of modules to improve system performance, Data

Manager( which aggregates sensory data) and Intelligent Agent( which
exploits the difference between consistency and timeliness of sensory data
in order to optimize query processing). QueryAgent provides two APIs,
turnon(QID) and turnoff(QID), which turn the sensor nodes on and off to
dynamically control energy consumption.

Data-Related Functions
WSN is a data-centric device, middleware would contain data management
functions such as data dissemination, data compression, and data storage.
i. Data Dissemination :
In WSNs, the sensor nodes deployed produce data. The data sensed
need to be transmitted to some special node or a sink for further
analysis, management, and control. Therefore, a data dissemination
protocol is required to provide effective data transmission from sensor

Divya T M, Assistant Professor, Dept.of ECE

 IOT & WIRELESS SENSOR NETWORKS 21EC741

nodes to the sink. Data dissemination protocols have a certain relation

to the routing protocols. The routing protocols are general and are
designed to find a path between the source and destination nodes. On
the other hand, data dissemination protocols should guarantee
successful transmission from nodes to the sink.
Data dissemination protocols consist of at least two phases:
1. triggering data transmission : This is the initial phase often
initiated by the sink, by sending out a query to inform sensor
nodes. The query contains information to guide data
transmission from the node to the sink, the frequency of data
reporting, the duration of interval in which data reporting
should take place, and so on.
2. Data transmission phase : sensor nodes report data to the
sink. Data dissemination protocols need to indicate whether
the data are to be transmitted in broadcast or unicast mode.
The protocols include Directed Diffusion (DD), two-tier data
dissemination (TTDD), sinks accessing data from
environments (SAFE).
Directed Diffusion (DD): The query is flooded. The initial data
are also broadcast to all neighbors to set up a reinforced
path, but subsequent data are transmitted only on the
reinforced path.
TTDD : sensor nodes need to announce the process to build
a grid structure. Then the query is flooded only in an area
smaller than a grid cell in order to find a nearby
dissemination node. The dissemination node is defined as the
node closest to the crossing point of the grid.
SAFE: SAFE avoids duplicate data and therefore conserves
energy.
ii. Data Compression :
The data compression techniques is important as the communication
components consumes most of the energy in WSNs. The data
compression techniques are important for the following reasons:
1. The data collected in neighboring sensor nodes are correlated,
especially when the deployment of sensor nodes is quite dense in
the network.
2. The occurrence of an event may be assimilated with a continuous-
time but random process, and sampling of the random processes
helps extract information content from the process.
The Compression techniques include the following :
1. Information theoretic–based techniques such as distributed
source coding using syndromes (DISCUS), This is a distributed
compression scheme for a dense microsensor network. WSNs
consist of sensor nodes in a treelike topology where the root is

Divya T M, Assistant Professor, Dept.of ECE

 IOT & WIRELESS SENSOR NETWORKS 21EC741

the sink, information is compressed or encoded at each node

incorporating the correlation with data from its parent node.
2. Data aggregation–based compression schemes such as tiny
aggregation service for ad hoc sensor networks (TAG). TAG
realizes several semantic-based aggregations such as MIN, MAX,
and SUM.
3. Sampling of a random process. If an application tolerates a
certain level of error, sensor nodes can adaptively reduce
sampling frequency.
iii. Data Storage :
Sensor nodes collect data related to the sensed events. Data need to
be stored, usually for future use. Several questions that merit
consideration for data storage are: What type of data need to be stored?
Where should the data be stored? How and for how long should data
be stored? The answers to these questions define the data storage
requirements of WSNs. There are two types of data in WSNs: the raw
data collected by the sensor nodes, and the results analyzed from the
data collected initially, such as from an event and its location.
The data storage schemes proposed are :
1. External storage (ES) : The data sensed are transmitted to an
external (centralized) host for storage. This approach is not energy-
efficient, since while all the data are hauled to a center, not all the
data are required for future query.
2. Local storage (LS). The data collected are stored locally in the sensor
node itself. Although the LS scheme is more energy-efficient than
ES, it is not efficient for query. For example, if frequent querying
occurs for data in a distant sensor node, LS consumes more energy
than if the data were centrally stored. An advantage of LS is that
the data location will be known in the query process.
3. Data-centric storage (DCS): the event data are stored based on their
event type and at some special ‘‘home nodes’’ that may not be the
original location of the data collected. Therefore, in DCS, a query
can be routed to the corresponding home nodes according to the
data type. This approach can be energy efficient.
4. Provenance-aware data storage (PADS): The event data are stored
locally in PADS, while the index or pointer of the data is stored in
some home hosts. PADS emphasizes the necessity of being able to
query the provenance of data for certain applications.
5. Multiresolution storage (MRS) : data are decomposed and classified
into levels: for example, level 0 for raw data, level 1 for finer data,
and level 2 for the coarsest data. Data of different levels will be
stored for different time durations. The level 2 data will be stored
for the longest term and raw data will be stored for the shortest.

Divya T M, Assistant Professor, Dept.of ECE

 IOT & WIRELESS SENSOR NETWORKS 21EC741

Performance and Traffic Management

Design factors for WSN
The performance of WNSs is affected strongly by other parameters in two
groups: basic models and network models. Basic models form the elementary
blocks based on which the network models can be analyzed.
WSNs usually consist of hundreds or thousands of sensor nodes scattered in
a geographical area and one or multiple sink(s) collecting information and
transmitting it through wireless channel.

Fig : WSN
sensor nodes need small simplified operating systems and energy-efficient
communication protocols.
The new metrics of Performance modeling and evaluation for WSNs are system
lifetime and energy efficiency, and the introduction of new traffic attributes.
1. These sensor nodes have different levels of resources within them, but
they all contain at least the following physical units: a radio unit with
a transceiver, a processing unit with a microcontroller and a memory,
a sensing unit with a sensor (or multiple sensors), and a power supply
unit, usually with a battery. Sensor nodes may have an additional unit
to support mobility or be equipped with a GPS-based unit. With the
development of micro electromechanical systems (MEMSs) and the new
battery technologies, sensor nodes might be able to carry more
resources. However, the resources of sensor nodes are still constrained
compared to the practically unconstrained physical and networking
interfaces and other resources in traditional network nodes. These
constraints have a direct impact on system and protocol design.
2. WSNs usually have a multihop physical topology. In the multihop two
tier architecture,The first type organizes all sensor nodes in a
hierarchical structure , here the sensor nodes at the first layer perform
only sensing, the sensor nodes in the second layer perform sensing and
data relaying, and so on. This topology can result in more efficient

Divya T M, Assistant Professor, Dept.of ECE

 IOT & WIRELESS SENSOR NETWORKS 21EC741

routing, but the topology formation is an energy-consuming task and

also increases the complexity of sensor nodes. A sensor node will
become dysfunctional if its energy is depleted and recharging is not
possible. The topology is usually variable and has multiple paths from
the source nodes to the sink. These attributes influence the design of
routing protocols in WSNs.
3. The traffic in WSNs flows starlike from sensor nodes to the sink. If there
are multiple sinks, multiple traffic flows will be generated between
sensor nodes and the sink. The sensor nodes gather data and report to
the sink according to the preconfigured rules. This many-to-one traffic
flow is called convergecast. which means many-to-one traffic flow from
sensor nodes to the sink. Therefore, sensor nodes closer to the sink
have the heavier burden for relaying, and due to higher energy
consumption they might become dysfunctional sooner.
The other design factors influence WSN are
4. Node deployment -Random, manual, one-time, iterative
5. Mobility - Immobile, partly, all; occasional, continuous; active, passive
6. Network topology- Single-hop, star, networked stars, tree, graph
7. Coverage- Sparse, dense, redundant
8. Connectivity- Connected, intermittent, sporadic
9. Network size- Hundred, thousand.
10. Communications- Laser, infrared, radio-frequency (narrowband,
spread spectrum, UWB)
WSN DESIGN ISSUES
This section includes design issues of WSN in
1. MAC protocols
2. Routing protocols
3. Transport protocols
1. MAC protocols :
MAC protocols affect the efficiency and reliability of hop-by-hop data
transmission. Existing MAC protocols such as the IEEE 802 series
standard may not be completely suitable for WSNs because of energy
efficiency.
MAC protocols can result in a waste of energy in the following ways:
i. A wireless channel is shared in a distributed manner, packet
collision cannot be avoided. The collided packets require
retransmission and result in energy waste.
ii. Most distributed wireless MAC protocols require control messages
for data transmission (e.g., request-to-send/clear-to-send in the

Divya T M, Assistant Professor, Dept.of ECE

 IOT & WIRELESS SENSOR NETWORKS 21EC741

IEEE 802.11 distributed coordination function). Control messages

consume energy.
iii. Overhearing and idle listening can also result in energy waste.
Overhearing means that a node receives packets destined for other
nodes. Idle listening refers to a situation where nodes there need to
listen on the channel to get its status.
MAC protocols for wireless sensor networks emphasize energy efficiency
through design of effective and practical approaches to deal with the
foregoing problems. For example, S-MAC designs an adaptive algorithm to
let sensor nodes sleep at a certain time.
Typical parameters used to measure performance of MAC protocols include
collision probability, control overhead, delay, and throughput.

2. Routing Protocols:
The routing protocols in WSNs are for setting up one or more path(s) from
sensor nodes to the sink. Since sensor nodes have limited resources,
routing protocols should have a small overhead. Therefore, the traditional
address-centric routing protocols for Internet (e.g., the routing information
protocol, open shortest path first, border gateway protocol) do not meet the
requirements of WSNs.
Data-centric routing is more suitable for WSNs because it can be deployed
easily, and due to data aggregation, it saves energy. routing protocols for
WSNs employ certain technique to minimize energy consumption e.g., data
aggregation and in-network processing & clustering. Directed Diffusion is
a data-centric routing scheme with three phases in its operation:
i. A sink broadcasts its interest across the network in query messages
with a special query semantic at a low rate.
ii. All the nodes cache the interest. When a node senses that an event
matches the interest, it sends the data relevant to the event to all
the interested nodes. Sink will also get the initial data and
‘‘reinforce’’ one of source nodes by resending the interest at a higher
rate.
iii. After the reinforcement propagation, the source nodes send data
directly on the reinforced path.
The performance of a routing protocol can be expressed through such
measures as computational overhead, communications overhead, path
reliability, path length, convergence rate, and stability.

3. Transport protocols :

Divya T M, Assistant Professor, Dept.of ECE

 IOT & WIRELESS SENSOR NETWORKS 21EC741

The transport protocols should handle congestion as the data streams are
convergent toward the sink, congestion is likely to occur at nodes around
the sink. Therefore, transport protocols should have mechanisms for loss
recovery; to guarantee reliability, mechanisms such as ACK and selective
ACK used in the TCP would be helpful. At the same time, reliability where
correct transmission of every packet should also be guaranteed.
The transport protocols should offer the following :
i. A hop-by-hop mechanism for congestion control - The hop-by-hop
mechanism can also lower the buffer requirement at the
intermediate nodes
ii. loss recovery
iii. Reduced packet loss & conservation of energy .
iv. Transport control protocols for WSNs should also avoid packet loss
as much as possible since packet loss translates to waste of energy.
v. it should guarantee fairness so that individual nodes can achieve
their fair throughput.
PERFORMANCE MODELING OF WSNs
The different performance measures are required to evaluate WSN.
The following are:
1. System lifetime:
(a) the duration of time until some node depletes all its energy.
(b) the duration of time until the QoS of applications cannot be
guaranteed.
(c) the duration of time until the network has been disjoined.
2. Energy efficiency:
Energy efficiency means the number of packets that can be
transmitted successfully using a unit of energy. Packet collision
at the MAC layer, routing overhead, packet loss, and packet
retransmission reduce energy efficiency.
3. Reliability: In WSNs, the event reliability is used as a measure to
show how reliable the sensed event can be reported to the sink.
For applications that can tolerate packet loss, reliability can be
defined as the ratio of successfully received packets over the total
number of packets transmitted.
4. Coverage: The coverage is defined as the ratio of the monitored
space to the entire space. Full coverage by a sensor network
means the entire space that can be monitored by the sensor
nodes.
5. Connectivity: For multihop WSNs, it is possible that the network
becomes disjointed because some nodes become dysfunctional.
The connectivity metric can be used to evaluate how well the

Divya T M, Assistant Professor, Dept.of ECE

 IOT & WIRELESS SENSOR NETWORKS 21EC741

network is connected and/or how many nodes have been

isolated.
6. QoS metrics : QoS requirements such as delay, loss ratio, and
bandwidth.

Basic Models
1. Traffic Model :
The four traffic models are used in WSNs:
a. Event-Based Delivery.
b. Continuous Delivery.
c. Query-Based Delivery.
d. Hybrid Delivery.

a. Event-Based Delivery :
i. sensor nodes monitor the occurrence of events passively and
continuously. When an event occurs, the sensor node begins
to report the event, and possibly an associated value, to the
sink. When delivering event data to the sink, a routing protocol
is often triggered in order to find a path to the sink. This
routing method is called routing on-demand.
ii. If an event appears frequently, at a node or a group of nodes,
the routing function is executed frequently, which results in
more energy consumption.
iii. An adaptive routing protocol may be required to set up a path
dynamically in advance if events occur frequently; otherwise,
the path is set up on-demand.
b. Continuous Delivery :
i. The data collected by the sensors need to be reported
regularly, perhaps continuously, or periodically
c. Query-Based Delivery:
i. The sink will issue query messages to sensor nodes to get the
up-to-date value for the information.
ii. Query messages may also carry a command from the sink to
the sensors about the information, reporting frequency and
other parameters of interest to the sink.
iii. In this delivery model, the sink broadcasts the query message,
a path is constructed automatically when the query arrives at
the sensor nodes, and the sensor nodes report their findings
according to the request in the query message.
d. Hybrid Delivery :
i. In some WSNs, the types of sensors and the data they sense
may be very diverse. For example, data may be reported
continuously by some nodes.
2. Energy Model:

Divya T M, Assistant Professor, Dept.of ECE

 IOT & WIRELESS SENSOR NETWORKS 21EC741

The radio communication function of sensor nodes is the most energy-

intensive function in the node.
There are two approaches to reducing consumption for sensor
communications.
i. The first approach is to design a communication scheme that
conserves energy inherently: for example, turning off the
transceiver for a period of time.
ii. The second approach is to reduce the volume of communications
through in-network processing.(These would entail functions
such as data aggregation and data compression since
computation tasks usually require less energy than do
communication tasks.)
Model for Sensing : The least amount of energy is consumed for
sensing. Let the sensing range be r. It can be assumed that the
power consumed to perform sensing over a circle with radius r is
proportional to 𝑟 2 or 𝑟 4 .
Model for Communication :
The energy for transmitting ‘l’-bit data over a distance ‘d’ is
𝐸𝑡𝑥 (𝑙, 𝑑) and the energy for receiving l-bit data over a distance d is
𝐸𝑡𝑥 (𝑙, 𝑑) = l𝐸𝑐 + led
𝐸𝑡𝑥 (𝑙, 𝑑) = 𝑙𝐸𝑐
𝑤ℎ𝑒𝑟𝑒 𝑒 = 𝑒1 ; s=2, d < 𝑑𝑐𝑟
𝑤ℎ𝑒𝑟𝑒 𝑒 = 𝑒2 ; s=4 , d > 𝑑𝑐𝑟
𝑤ℎ𝑒𝑟𝑒 𝐸𝑐 𝑖𝑠 𝑡ℎ𝑒 𝑏𝑎𝑠𝑒 𝑒𝑛𝑒𝑟𝑔𝑦 𝑟𝑒𝑞𝑢𝑖𝑟𝑒𝑑 𝑡𝑜 𝑟𝑢𝑛 𝑡ℎ𝑒 𝑡𝑟𝑎𝑛𝑠𝑚𝑖𝑡𝑡𝑒𝑟 𝑜𝑟 𝑟𝑒𝑐𝑖𝑒𝑣𝑒𝑟 𝑐𝑖𝑟𝑐𝑢𝑖𝑡𝑟𝑦.
The optimal distance between relay nodes 𝑑𝑚 is computed as
follows.

𝑠 𝑎1
𝑑𝑚 = √
𝑎2(𝑠−1)
𝑤ℎ𝑒𝑟𝑒 𝑎1 = 2𝐸𝑐 , 𝑎2 = 𝑒1 𝑜𝑟 𝑒2
Model for communication:
i. A sensor node usually has a microcontroller or microCPU
performing computations. Low power means that the CPU
consumes low energy per clock cycle.
3. Node Model:
i. To conserve energy, a common approach is to let nodes sleep
when they have no need to transmit or receive.
ii. The sensor nodes have two states: active (A) and sleep (S). The
length of the active and sleep period are geometrically distributed
random variables with a mean value of p and q time slots,
respectively. The active phase is divided further into R and N
states. In the R state, sensor nodes can transmit or receive data
and/or generate data according to a Poisson process with an

Divya T M, Assistant Professor, Dept.of ECE

 IOT & WIRELESS SENSOR NETWORKS 21EC741

average rate of g. In the N state, nodes can transmit data only if

there are packets backlogged in their buffer.
iii. Two state Discrete-time Markov chain is used for the next-hop
nodes. The two states defined for the next-hop node are wait (W)
and forwarding (F). State W means that all the next-hop nodes
are in either the S or N state and unable to receive data from the
node in mind. F represents that there is at least one next-hop
node.

Fig : Discrete-time Markov chain

Network Models :
1. MAC Model
2. Routing Model
3. System Model
1. MAC Model :
The channel access and the channel is controlled and
allocated by MAC protocols. If packet collision occur on
the channel and should be handled by the MAC. MAC
protocol should ensure that the data are transmitted
successfully in a time slot.
The successful transmission can be defined as, Let r be
the common maximum radio range. The one-hop
transmission from transmitter i to the receiver j is
successful if
(1) the distance between i and j should be less than or
equal to r. i.e
𝑑𝑖,𝑗 ≥ 𝑟
(2) for every other node, k, which is also receiving at the
same time,
𝑑𝑖,𝑘 > 𝑟
(3) for every other node, l, which is transmitting at the
same time,
𝑑𝑙,𝑗 > 𝑟
WSN employs a CSMA/CA mechanism with
handshaking as in an IEEE 802.11 DCF. The

Divya T M, Assistant Professor, Dept.of ECE

 IOT & WIRELESS SENSOR NETWORKS 21EC741

performance metrics such as throughput, delay, and

collision probability can be obtained for the given
number of competing nodes.
2. Routing Model :
The energy consumed for a generic route P can be computed as
follows:
𝐸(𝑃) = ∑ 𝐸𝑖,𝑛𝑝 (𝑖) (𝑙𝑖 , 𝑑𝑖 )
𝑖
𝑤ℎ𝑒𝑟𝑒 𝑛𝑝 (𝑖) 𝑖𝑠 𝑡ℎ𝑒 𝑛𝑒𝑥𝑡 ℎ𝑜𝑝 𝑖 𝑜𝑛 𝑝𝑎𝑡ℎ 𝑃.
𝐸𝑖,𝑛𝑝 (𝑖) (𝑙, 𝑑)𝑖𝑠 𝑡ℎ𝑒 𝑒𝑛𝑒𝑟𝑔𝑦 from node i to node 𝑛𝑝 (𝑖)

The advantages of data-centric routing over address-centric routing in

supporting data aggregation were found as follows :
1. If the diameter of the set of source nodes (X) is shorter than the
minimal length of the shortest path from any source node to the sink
(Dmin), the total number of transmissions under data-centric routing
is smaller than with address-centric routing. Therefore, data-centric
routing is more energy efficient.
2. The larger the distance between X and Dmin, the more energy is
conserved by the data-centric routing.
3. System Model :
A closed-loop model has been constructed to consider the
sensor node model, MAC protocol, and routing policy all at the
same time. This model consists of three submodels. Sensor
node model, interference model and network model.

Fig: closed loop model for the system

The system performance metrics such as average energy
consumption and average delay have also been calculated using
this closed loop model.
1. Construct a DTMC sensor node model : Each sensor node i
with distribution pi and the probability that data are received
in a time slot in node I is computed.

Divya T M, Assistant Professor, Dept.of ECE

 IOT & WIRELESS SENSOR NETWORKS 21EC741

2. Solve the network model using queuing network analysis to

calculate the average data transmission rate between any
pair of sensor nodes and the average throughput for each
sensor node.
3. Given the average data transmission rate as input to the
interference model, compute the value of the probability that
data are transmitted successfully in a time slot in a node i(𝑏𝑖 )
4. 𝑏𝑖 is used as input to the sensor node model iterating
through steps 1 to 3.

******

Divya T M, Assistant Professor, Dept.of ECE